viplismism / rlm-cli

What is rlm-cli?

rlm-cli is a TypeScript CLI that brings Recursive Language Models to your terminal, letting LLMs process huge contexts by generating Python code for chunking, sub-queries, and aggregation—instead of cramming everything into one prompt. Fire up `rlm` for an interactive REPL where you load files, recursive directories (skipping node_modules like aws cli recursive copy), globs, or URLs, then query with Anthropic, OpenAI, or Google models. It saves full trajectories for review and includes benchmarks on long-context datasets.

Why is it gaining traction?

Unlike basic LLM CLIs, rlm-cli shows the full loop live—code gen, execution, sub-queries—in a polished TUI with trajectory viewer, beating direct prompts on tasks like timeline counting or narrative QA. Recursive glob loading (cli recursive search) and single-shot scripting (rlm run --file codebase/ "find bugs") make it dead simple for big repos, while multi-provider switching and GitHub-friendly setup hook devs tired of context blowups. Benchmarks prove it handles 10MB+ without losing accuracy.

Who should use this?

Backend devs dissecting monorepos or legacy codebases, where cli find recursive falls short on semantic search. AI tinkerers benchmarking long-context perf, or ops folks scripting analysis on logs/docs via stdin/URLs. Pairs well with GitHub Copilot for code review, or GitHub Actions for automated repo audits on Linux/Ubuntu/Windows.

Verdict

Grab it if you're hitting LLM context walls—installs clean via npm, docs are crisp, interactive mode shines. At 66 stars and 1.0% credibility, it's early but mature enough for daily use; watch for more benchmarks as it grows.

(198 words)